Cardiac Alternans: From Bedside to Bench and Back

Abstract

Cardiac alternans arises from dynamical instabilities in the electrical and calcium cycling systems of the heart, and often precedes ventricular arrhythmias and sudden cardiac death. In this review, we integrate clinical observations with theory and experiment to paint a holistic portrait of cardiac alternans: the underlying mechanisms, arrhythmic manifestations and electrocardiographic signatures. We first summarize the cellular and tissue mechanisms of alternans that have been demonstrated both theoretically and experimentally, including 3 voltage-driven and 2 calcium-driven alternans mechanisms. Based on experimental and simulation results, we describe their relevance to mechanisms of arrhythmogenesis under different disease conditions, and their link to electrocardiographic characteristics of alternans observed in patients. Our major conclusion is that alternans is not only a predictor, but also a causal mechanism of potentially lethal ventricular and atrial arrhythmias across the full spectrum of arrhythmia mechanisms that culminate in functional reentry, although less important for anatomic reentry and focal arrhythmias.

In 1872, German physician Ludwig Traube published his observations on a patient with alcoholic cardiomyopathy and heart failure whose pulse alternated in intensity from beat to beat (pulsus alternans), presaging his sudden death only 2 months later¹. After the invention of the electrocardiogram, Herring in 1909 reported alternans of both the QRS complex (QRSA) and T-wave (TWA) in animal experiments,² and established a link between electrical and mechanical (pulsus) alternans, confirmed in humans a year later by Lewis.³ Over the past 150 years, mechanical pulsus alternans and electrical alternans have been observed in many settings in which arrhythmias are common, including heart failure, acute and chronic myocardial ischemia, cardiomyopathy, inherited arrhythmia syndromes, and drug and electrolyte disturbances. Extensive clinical, experimental, and theoretical studies have been carried out to elucidate the mechanisms of alternans and its links to arrhythmias and sudden cardiac death^4–14 (searching PubMed with “alternans, cardiac” returned 2329 articles dating from 1947 to 2022). In the 1990s, human clinical trials conclusively established the link between electrocardiographic TWA and arrhythmia risk.^13,15,16 In long QT syndromes (LQTS), TWA prevalence and arrhythmia risk increase in parallel with the corrected QT intervals (Figure 1A).¹⁷ Moreover, clinical observations suggest that TWA is more than just a predictor of arrhythmia risk, but may also be a causal mechanism. For example, there are many published examples from LQTS patients in which TWA immediately precedes the onset of its signature arrhythmia, Torsade de pointes (Figure 1 B and C).^18–23 On the other hand, many episodes of Torsade de pointes also occur without immediately preceding TWA, often following a short-long-short sequence (Figure 1D).^24–26 Today, macroscopic TWA is used in assessing arrhythmia risk in patients with LQTS,²⁷ and microvolt TWA test is used for risk stratification for patients with heart failure.^28,29

Figure 1.

T-wave alternans (TWA) and arrhythmogenesis.

A, TWA prevalence and arrhythmias in long QT syndrome (LQTS) patients.¹⁷ Left, TWA prevalence vs corrected QT interval (QTc). Right, Cardiac event rate vs QTc with (black) and without (white) TWA. B, TWA transitioning directly to Torsade de pointes (TdP) without a clear preceding premature ventricular complex (PVC).¹¹ C, TWA in which TdP is initiated by a PVC falling on the large T-wave of the alternating beats.²³ D, TWA, followed sometime later by TdP initiated by a short-long-short sequence.²⁴

In this review article, our goal is to provide a holistic portrait of cardiac alternans and its relationship to cardiac arrhythmias by linking the relevant clinical observations to experiments and theory, focusing primarily on life-threatening ventricular arrhythmias but also including atrial arrhythmias. We begin by summarizing the cellular and tissue mechanisms of alternans that have been demonstrated theoretically and experimentally, including voltage-driven and calcium (Ca)-driven mechanisms, and their involvement in the dynamics of arrhythmia initiation. We link these mechanisms to the electrocardiogram characteristics of alternans observed in patients under various pathophysiological conditions. Based on experimental and simulation results, we discuss the relevance of alternans to arrhythmogenesis across a spectrum of disease states, distinguishing between alternans as a predictor versus a direct causal factor. We conclude by discussing alternans in risk stratification and as a therapeutic target for preventative therapies.

Cellular Mechanisms of Alternans

As an excitable medium, the heart is a dynamical system capable of exhibiting a wide range of spontaneously self-organizing behaviors, from pacemaking driving the normal heartbeat to pathophysiological phenomena such as triggered activity and reentry. Alternans falls within this spectrum of behaviors, and can be precipitated by a variety of conditions, including changes in heart rate, genetic or post-translational changes in ionic current conductances or kinetics, altered intracellular Ca loading and sarcoplasmic reticulum (SR) Ca release properties, etc. In this article, we use the term repolarization alternans to refer to both TWA and action potential duration (APD) alternans, since APD is the major determinant of the T-wave and QT interval. Likewise, we use the term mechanical alternans to refer equivalently to pulsus alternans and intracellular Ca transient (CaT) alternans, since the CaT is the major determinant of contractile strength.

From a dynamical systems viewpoint, cardiac arrhythmias can be classified into 3 major categories³⁰: Ca cycling disorders (typified by catecholaminergic polymorphic ventricular tachycardia [CPVT]), prolonged repolarization disorders (typified by LQTS), and early repolarization disorders (typified by Brugada and short QT syndromes). As opposed to inherited arrhythmia syndromes, the more common complex diseases associated with lethal ventricular tachycardia (VT) and ventricular fibrillation generally involve more than a single category. For example, chronic heart failure involves both abnormal Ca cycling and prolonged repolarization, and acute ischemia involves both abnormal Ca cycling and early repolarization. Alternans is a feature common to all 3 arrhythmia categories.

The first rigorous theory of repolarization alternans was developed in 1968 by Nolasco and Dahlen³¹ who showed that in frog heart, the occurrence of APD alternans is determined by the slope of the APD restitution (APDR) curve. Since then, new theories and mechanisms of alternans have been uncovered through rigorous theoretical analysis confirmed by computer simulations and supported by experimental and clinical observations. Alternans can be generally subdivided into voltage-driven mechanisms and Ca-driven mechanisms, related to the fact that the heart contains 2 excitable subsystems, the voltage subsystem supporting electrical wave propagation through cardiac tissue and the Ca-induced Ca release subsystem supporting Ca wave propagation within cardiac cells. Each subsystem is capable of generating alternans and more complex dynamical phenomena such as spiral/scroll wave reentry on its own. In voltage-driven alternans, APD dynamics drive repolarization alternans with mechanical alternans following passively, reflecting the fact that the APD is an important factor regulating CaT amplitude. In Ca-driven alternans, intracellular Ca dynamics drive mechanical alternans and repolarization alternans follows passively, since the CaT amplitude feeds back to influence APD via its effects on Ca-sensitive ionic conductances such as the L-type Ca current, electrogenic Na-Ca exchanger, and Ca-activated currents. However, since the voltage and Ca subsystems are inherently bidirectionally coupled via the effects of intracellular Ca on ionic conductances controlling membrane voltage and vice versa, the 2 subsystems always exert joint influences on each other, which can be synergistic or antagonistic depending on underlying conditions. From a practical standpoint, this means that when both voltage and Ca subsystems are free-running, alternans is never purely voltage-driven or purely Ca-driven; only their relative predominance varies.

Voltage-Driven Alternans

Voltage-driven alternans can be classified into 3 sub-mechanisms^12,32: (1) alternans driven by a steep APDR slope at fast heart rates or short diastolic intervals (DI), as originally described by Nolasco and Dahlen³¹; (2) alternans driven by factors prolonging repolarization such as early afterdepolarizations (EAD) in LQTS; and (3) alternans driven by all-or-none early repolarization as in Brugada and short QT syndromes. In all 3 cases, the features of the APDR curve play a critical role (Figure 2).

Figure 2.

Voltage-driven alternans. A, Steep action potential duration restitution (APDR)-slope–induced action potential duration (APD) alternans. B, Early afterdepolarization (EAD)-induced APD alternans. C, Early repolarization-induced APD alternans. Left, Experimental action potential (AP) recordings for the 3 cases, obtained from frog ventricle by Nolasco and Dhalen (A),³¹ a rabbit ventricular myocyte with hypokalemia-induced EADs (B),⁴⁵ and canine epicardial ventricular muscle during simulated ischemia-induced early repolarization (C).⁴⁹ Middle, APDR curves for the 3 cases obtained from corresponding computer simulations.^12,32,47 The intersection of the APDR curve (black) and pacing cycle length (PCL) line (blue) is the equilibrium point (open circle). If the slope of the APDR curve at this point is >1, the equilibrium point is unstable. Using the graphical cobweb approach pioneered by Nolasco and Dhalen,³¹ starting from an initial APD and diastolic interval (DI; star symbol), the subsequent APD and DI values alternate in a growing pattern (red dashed lines and arrows), eventually reaching steady-state APD alternans (green dashed box and colored circles). Right, Corresponding plots of APD versus PCL for the 3 cases.^12,32,47 For each PCL, more than 20 consecutive APDs are superimposed. Two APD values at the same PCL indicate alternans, and many APD values indicate high periodicity or chaos.

Steep APDR slope-induced APD alternans

The APDR curve can be measured experimentally by the extrastimulus method (ie, pacing heart tissue at a fixed pacing interval [S1S1] and introducing timed extrastimuli [S1S2] at different DIs), or by the rapid pacing method (ie, pacing the heart tissue at incrementally faster rates to shorten the DI until 1:1 capture is lost). In both cases, the APDR curve is then obtained by plotting APD as a function of the preceding DI. The resulting APDR curve represents the collective behavior of the recovery kinetics of the ionic currents that are active during the action potential. Depending on its recovery time constants, each ionic current contributes to a different DI range.³³ Na channels recover quickly and thus affect APDR at short DIs (although their recovery kinetics are slowed by acute ischemia due to the depolarized resting potential, Na channel remodeling or drugs). L-type Ca channels (LCCs) recover more slowly than Na channels, affecting APDR at short to intermediate DIs, and time-dependent K channels recover even more slowly, affecting ADPR over intermediate to long DIs. In general, APD prolongs monotonically as DI increases (Figure 2A, middle), but non-monotonic responses can also occur.³⁴

The ADPR curve can be represented mathematically by the equation

$$AP{D}_{n+1}=f(D{I}_n)$$ (1)

such that once the function $f(DI)$ is obtained from the experimental data, one can predict $AP{D}_{n+1}$ for any $D{I}_n.$ In 1968, Nolasco and Dahlen³¹ used a graphic representation of this iterative process to show that persistent APD alternans developed in frog hearts when the slope of the APDR curve exceeded 1 at the point of its intersection with the pacing cycle length line (Figure 2A, middle).¹²

Since the APDR slope generally becomes steeper as heart rate increases, this mechanism of alternans is generally promoted by fast heart rates (but can occur at normal or slow heart rates when APD is long enough so that the DI is short, as illustrated by the example in the left of Figure 2A). This elegant analysis in the frog heart, however, has only partial relevance to the mammalian heart, for 2 reasons. First, the frog heart has a very limited intracellular Ca cycling system due to its rudimentary SR compared with mammalian hearts.³⁵ In normal mammalian heart, however, Ca-driven alternans plays a much greater role in initiating APD alternans as heart rate increases, which typically occurs well before the APDR slope exceeds one in experiments.^36–39 Second, the APDR slope >1 criterion is strictly valid only if APD is solely a function of the previous DI, as indicated by Equation 1. If short-term memory is present (due to ionic currents with very slow kinetic components or slow heart rate-dependent intracellular ion accumulation/depletion kinetics) such that the APD is also influenced by the pacing history prior to the last DI, then the APDR slope >1 criterion for instability no longer holds. This is the case in mammalian hearts, since the APDR curve obtained by the extrastimulus method differs when the baseline pacing rate is changed, and also differs from that obtained by the rapid pacing method.^34,37,40 Thus, in the presence of short-term memory, the APDR curve is not unique, and depends on the pacing protocol and history.^34,37,40 Theoretically, it is possible to include the short-term memory effects to assess more precisely the role of the APDR on the genesis of APD alternans, and when done,⁴¹ an even steeper APDR slope is generally required for alternans to occur.

EAD-induced alternans

EADs are an all-or-none phenomenon associated with APD prolongation due to reduced repolarization reserve,^42,43 and at some heart rates exhibit an alternating pattern in which an EAD prolongs the APD of every other beat (Figure 2B, left).^20,44,45 The transition to alternans is related to the sudden sharp jump in the APDR curve typically occurring at long DIs due to the all-or-none occurrence of EADs (Figure 2B, middle). The dynamical mechanism of alternans can still be illustrated by the same graphical method. However, the APDR slope >1 criterion for the onset of alternans is not strictly valid because short-term memory effects also play a key role, which can be mathematically solved using iterative maps.^12,46 Differing from steep APDR slope-induced alternans, EAD-induced alternans is generally promoted by normal or slow heart rates and suppressed by fast heart rates (Figure 2B, right).^32,47 Theoretically, APD alternans can also occur when APD is very long without the presence of EADs, induced by steep APDR at normal or slow heart rates with either very short DIs as in Figure 2A or long DIs.⁴⁸

Early repolarization-induced alternans

In regions of the heart with a high density of transient outward K current (I_to) or other I_to-like currents, the action potential often takes on a spike-and-dome appearance that is sensitive to all-or-none repolarization in which the action potential dome suddenly disappears, leaving only the initial spike with a very short APD (Figure 2C, left). Similar to the EAD case, APD alternans can be illustrated graphically as arising from the steep APDR slope reflecting the interaction between I_to-induced early repolarization and the slow recovery kinetics of the slow delayed rectifier K current⁴⁶ (Figure 2C, middle). As with EAD-induced alternans, however, rigorous analysis of the onset of alternans requires iterative map approaches integrating the short-term memory effects.⁴⁶ Early repolarization-induced alternans has been demonstrated experimentally in mammalian ventricular tissue^49–51 as well as computer simulations,^12,52,53 and typically occurs at normal to elevated heart rates (Figure 2C, right).

Complex behaviors and chaos

In all 3 voltage-driven alternans mechanisms described above, it is the steep regions of the APDR curve that play the key causal roles in generating APD alternans. Moreover, in all 3 cases, alternans is a precursor to more complex behaviors, such as higher order periodicity and chaos. In the steep APDR slope case, fast heart rates leading to loss of 1:1 capture can lead to n:m capture (ie, Wenckebach phenomenon) manifested in higher order periodicity and chaos.^12,54–56 In this case, complex APD behaviors only occur after 1:1 capture is lost (Figure 2A, right). In contrast, for both the EAD and early repolarization cases, higher order periodicity and chaos can occur without loss of 1:1 capture, as indicated by the regions with multiple APD values at the same pacing cycle length in the right of Figure 2B and 2C.^{32,44–47,49} The dynamical mechanisms of these complex behaviors have also been rigorously analyzed using iterative map approaches,^44,46 which show that short-term memory plays a key role. Thus, both alternans and more complex behaviors originate from the same APDR-induced instabilities, with the alternans being an early prognosticator of more complex instabilities to come.⁵⁷

Ca-Driven Alternans

Just as the heart is an excitable system designed to initiate a coordinated contraction/relaxation cycle by propagating the action potential from myocyte to myocyte, each myocyte’s internal Ca cycling system is an excitable subsystem in its own right, in which the Ca flowing through LCCs during the action potential activates the release of a large amount of intracellular Ca stored in the SR via Ca-induced Ca release (Figure 3A through 3C). Under the appropriate conditions, the CaT can exhibit spontaneous alternans on its own, even when the action potential waveform is voltage-clamped to prevent voltage-driven APD alternans from interfering (Figure 3D).^58–61

Figure 3.

Cardiac Ca cycling.

A, Schematic diagram of a single calcium release unit (CRU) at a T-tubular membrane-sarcoplasmic reticulum (SR) membrane junction. B, A high-resolution image of the Ryanodine receptor (RyR) clusters in individual CRUs located in the T-tubule network of a rat ventricular myocyte.⁶² C, Simulation demonstrating the Ca signaling hierarchy of quarks (q), sparks (s), and spark clusters (c) in a computer model of a ventricular myocyte.⁶⁶ In the time sequence illustrated, one of the clusters at the left eventually propagates as a mini-wave (w) by recruiting adjacent clusters. D, Primary Ca-driven alternans in a rabbit ventricular myocyte during pacing with a action potential (AP) clamp waveform.⁶⁷ E, The fractional SR Ca release curve measured in a rabbit ventricular myocyte.⁶⁹ F, Ca transient (CaT) restitution curves at 2 different extracellular [Ca]_o recorded from mouse ventricular myocytes using the S1S2 extrastimulus method.⁷² CaM indicates calmodulin; CaMKII, Ca/calmodulin-dependent protein kinase II; and NCX, Na-Ca exchange.

The mechanism of Ca-driven alternans is more complex than voltage-driven alternans, due to the complex spatial arrangement of the network of intracellular Ca release units (CRUs) within each myocyte’s cytoplasm (Figure 3B).⁶² Within the CRU network, each CRU consists of a cluster of Ca release channels called ryanodine receptors (RyRs) residing in the SR membrane that are closely apposed to a cluster of LCCs residing in invaginations of the surface membrane called T-tubules (Figure 3A). Together with a host of other regulatory proteins, they form a Ca-signaling microdomain at each of these T-tubular membrane/SR membrane junctions. When one or more of the LCCs in the microdomain opens during an action potential, the Ca ions flowing into the narrow intermembrane space cause the nearby RyRs to open and release Ca ions stored within the SR into the cytoplasm, which can be imaged confocally as Ca sparks (Figure 3C). As the number of LCCs that open increases, more CRUs release their local store of SR Ca into the cytoplasm and the resulting Ca sparks summate together in a graded fashion to generate the whole cell CaT that determines the strength of contraction. Subsequently, the released Ca is pumped back into the SR by sarco-endoplasmic reticulum Ca ATPase (SERCA) pumps in the SR membrane or extruded into the extracellular space via the Na-Ca exchanger in the surface membrane.

Several additional features of the CRU network play critical roles in the genesis of CaT alternans. The first feature is related to the fact that the CRUs in the network are coupled via Ca diffusion both inside the network SR and throughout the cytoplasm, such that the firing of a CRU affects and is affected by its neighbors. For example, when 1 CRU fires, the released Ca diffusing to the neighboring CRUs can cause their RyRs to open so that they too fire; that is, a firing CRU can sometimes recruit nearby CRUs to fire, a process called spark-induced sparks.⁶³ Therefore, besides individual sparks triggered directly by LCC openings, spark clusters (macrosparks) and spark-induced Ca waves can also occur via spark-induced sparks (Figure 3C).^64–66 In addition, as the SR Ca load increases, the greater amount of Ca released by each spark increases the recruitment probability of additional spark-induced sparks. This causes the fraction of SR Ca released by the CRU network to increase steeply when the SR Ca load exceeds a certain level.^67,68 The relationship between the fraction of SR Ca released and SR Ca load can be measured experimentally⁶⁹ and quantified as the fractional SR Ca release curve (Figure 3E). The second key feature is that once a CRU fires and releases its local SR Ca store, there is a time delay before it becomes available to fire again, called Ca spark restitution.^70,71 It is this recovery period that underlies the CaT restitution curve shown in Figure 3F.⁷² Since SR refilling is faster than RyR recovery,^70,73 the CaT restitution curve is mainly determined by RyR recovery kinetics, mediated by regions of the RyR protein that sense the intraluminal SR Ca levels, either directly⁷⁴ or through interactions with intraluminal accessory SR proteins such as triadin, junction, and calsequestrin.⁷⁵ With this background, it is now possible to understand the 2 main mechanisms that cause CaT alternans^{56,63,68,76–79} and how they synergize with each other at fast heart rates.⁶⁸

CaT alternans driven by SR Ca load alternans

Based on experimental studies^59,60,80 demonstrating that CaT alternans could be elicited at heart rates too slow to encounter RyR refractoriness and CaT restitution, Eisner et al⁷⁶ proposed that a steep fractional SR Ca release curve could explain CaT alternans in the following manner. In the steep region of the fractional SR Ca release curve, a small change in SR Ca load results in a larger change in fractional SR Ca release (due to more spark-induced sparks), creating a positive feedback relationship. Thus, if slightly more SR Ca is released on the n^th beat, the resulting CaT will be larger, but the SR Ca load becomes more depleted. If the SERCA pump is not capable of restoring all of the released Ca back into the SR, this will result in a smaller SR Ca load prior to the (n+1)^th beat. This in turn causes a smaller fractional Ca release and CaT on that beat, such that more SR Ca will be available for release on the (n+2)^th beat, resulting in a greater fractional SR Ca release and CaT, and so forth in an alternating fashion. This mechanism has been formulated mathematically using iterative map approaches,^10,56,68,77 and its prediction that CaT alternans is promoted by increasing the steepness of the fractional SR Ca release curve and decreasing the SR Ca reuptake rate via SERCA (Figure 4A) has been confirmed experimentally in isolated ventricular myocytes.⁸¹ Since spark-induced spark recruitment enhances the steepness of fractional SR Ca release curve,^67,68 this mechanism is also consistent with the observation that Ca mini-waves occur during CaT alternans (Figure 4B). Relevant to heart disease, computer simulations show that T-tubule disruption, a feature of heart failure remodeling, steepens the fractional SR Ca release curve by reducing LCC-triggered CRU firings to allow more recruitments to occur, and thereby promotes SR Ca load alternans driving CaT alternans.⁸²

Figure 4.

Mechanisms of Ca-driven alternans. A, Calcium transient (CaT) alternans driven by sarcoplasmic reticulum (SR) Ca load alternans. Plot of the slopes of the fractional SR Ca release curve versus sarco-endoplasmic reticular Ca ATPase (SERCA) pump rate, in which the line indicates the region above which CaT alternans occurs. CaT alternans is promoted by either reducing the SERCA pump rate or increasing the steepness of the fractional SR Ca release curve (arrows). B, Confocal linescans of Ca fluorescence showing Ca mini-waves during alternans recorded in a rat ventricular myocyte⁶⁰ (top) and simulated in a computer model (bottom).⁶³ C, Transition from microscopic CaT alternans to macroscopic alternans as pacing frequency was increased in a rat ventricular myocyte.⁹⁰ Left column, Confocal image and traces at the slower pacing rate show microscopic alternans (lower trace) at the site (marked by arrow) in the confocal image above, in the absence of global macroscopic alternans (upper traces). Right column, At the faster pacing rate, both microscopic (lower trace) and macroscopic alternans (upper traces) are present. Moreover, the 2 ends of the myocyte (superimposed black and gray traces in the middle trace) are alternating out-of-phase, forming subcellular discordant alternans.

CaT alternans driven by RyR refractoriness

In the CaT alternans mechanism above, it is obligatory for the SR Ca load to alternate concomitantly with the CaT, since the Ca release is uniquely determined by the SR Ca load. However, other experiments have shown that CaT alternans can occur without the SR Ca load alternating in the expected fashion,^83–85 indicating that another mechanism must also be at play. In experimental studies in both ventricular and atrial myocytes,^61,84–89 this other mechanism has been identified as RyR refractoriness, and explained by a complementary theory, called the 3R theory.^63,67 The “3R” refers to 3 fundamental phenomenological properties of CRUs: Random firings due stochastic LCC and RyR openings, Refractoriness requiring a recovery period for a CRU to regain excitability after firing, and Recruitment of neighboring unactivated CRUs to fire and initiate a spark cluster or Ca wave. CaT alternans without SR load alternans can occur during pacing when, due to RyR refractoriness, some CRUs fire only on every other beat (or even randomly).^63,90 If the CRUs are not strongly coupled, their firings occur randomly out-of-phase, and thus the summation of sparks gives rise to a whole-cell CaT without macroscopic alternans. However, if the CRUs are strongly coupled such that spark-induced spark recruitment occurs, the firings from adjacent CRUs can become locally synchronized, forming firing clusters or mini-waves (Figure 4B). Due to this recruitment or synchronization effect, as more CRUs fire in 1 beat and less in the next beat, whole cell CaT begins to alternate macroscopically. In other words, if there is no coupling between CRUs or the coupling is weak, one can see alternans in individual CRUs but not in the global signal. Once the coupling is strong enough, however, the individual alternans become synchronized to produce global alternans. This was directly observed during experiments in rat ventricular myocytes by Tian et al⁹⁰ who showed that microscopic alternans (ie, alternans in individual CRUs) but not macroscopic alternans occurred at slower pacing rates, whereas both microscopic and macroscopic alternans occurred together at faster pacing rates as coupling and recruitment became stronger with increased SR Ca loading (Figure 4C). Under such conditions, computer simulations^63,79 have shown that alternans can occur even when SR Ca load is artificially clamped at a constant level. However, since the spark firing rate and recruitment depend on the SR Ca load, SR Ca load still plays an important role in this mechanism of alternans.^68,79 Similarly, although SERCA activity is not an apparent parameter in this theory, it also plays an important role by affecting CRU coupling and the SR Ca load.⁷⁹

A unified theory of CaT alternans

In the 2 mechanisms of CaT alternans, one depends on encountering short DIs at which RyR are still partially refractory and the other does not. The refractoriness-dependent mechanism is promoted by fast heart rates, as necessary to encounter refractoriness, but the SR Ca load-dependent mechanism can occur at any heart rate as long as SR Ca loading conditions for alternans are met (Figure 4A). A unified theory has recently been developed that combines both mechanisms into the same theoretical framework.⁶⁸ This theory demonstrates that the 2 mechanisms can work in synergy to promote CaT alternans at fast heart rates encountering RyR refractoriness. Therefore, both mechanisms may be responsible for CaT alternans observed at fast heart rates.^84,88,91 At slow heart rates at which RyRs fully recover between beats, however, the SR Ca load-dependent alternans mechanism alone is responsible. Despite this difference, the 2 mechanisms are promoted by several common factors,^68,79 including disrupted T-tubules^82,92 and enhanced CRU coupling promoted by increased RyR sensitivity, reduced SERCA activity, reduced CRU spacing, and an appropriate SR Ca load. These factors promote recruitment to induce Ca mini-waves (Figure 4B) that play important roles in both mechanisms.^59,63,80,93 Among these factors, SERCA pump activity (which is highly regulated by catecholamines) is important to both mechanisms, supported by experiments that CaT alternans is promoted by SERCA reduction.^81,94–96 Most of the other common factors, such as T-tubule disruption,^97–99 reduced CRU spacing,¹⁰⁰ leaky RyRs, and reduced SERCA activity are also promoted by heart failure remodeling.

Subcellular discordant CaT alternans

The pattern of CaT alternans within a myocyte can become out-of-phase spatially, such that during the same beat, CaT alternates high-low in 1 region but low-high in an adjacent region. This so-called subcellular discordant CaT alternans (Figure 4C, trace labeled left/right) has been well documented in experimental studies.^59,101–104 Theoretical analyses combined with simulations^92,105,106 have shown that when Ca-to-APD coupling is negative (see next section), subcellular CaT alternans can occur whether the heart is paced with either a free-running action potential or a constant voltage-clamped waveform. Subcellular discordant CaT alternans does not occur when Ca-to-APD coupling is positive unless SR Ca release properties are heterogeneous¹⁰³ or the cell is paced with a constant voltage-clamped waveform.¹⁰⁶ Subcellular discordant CaT alternans is also promoted by T-tubule disruption or in cells with sparse T-tubule structures.⁹² When subcellular discordant alternans occurs, the beat-to-beat differences in whole-cell CaT amplitude are reduced (or even eliminated if exactly half of the myocyte is out-of-phase with the other half). The amplitude of APD alternans is also reduced, commensurate with the reduced amplitude of the whole-cell CaT alternans.

Coupling of Voltage and Ca

As noted previously, voltage and Ca are bidirectionally coupled via Ca-dependent ionic currents/transporters as well as by Ca-dependent signaling (Figure 3A). For example, a large CaT increases inward Na-Ca exchange current, tending to prolong APD, but also accelerates LCC inactivation and increases both the slow component of the delayed rectifier K current and the Ca-activated small conductance K current,^107,108 tending to shorten APD. If the net effect is to enhance inward current during the action potential plateau, the Ca-to-APD coupling relationship is said to be positive, such that increasing the CaT amplitude prolongs APD. Thus, during alternans, the larger CaT corresponds to the longer APD, referred to as electromechanically concordant alternans (Figure 5A).¹⁰⁹ If the net effect is to enhance outward current during the action potential plateau, Ca-to-APD coupling is said to be negative, such that increasing CaT amplitude shortens APD. In this case, during alternans the larger CaT corresponds to the shorter APD, referred to as electromechanically discordant alternans (Figure 5B).

Figure 5.

Effects of voltage and Ca coupling on alternans dynamics. A–D, Action potential duration (APD)-to-Ca coupling is assumed to be positive and labels refer to Ca-to-APD coupling sign. A, Electromechanically concordant alternans (EMCA) in a rabbit ventricular myocyte with positive Ca-to-APD coupling.¹⁰⁹ B, Electromechanically discordant alternans (EMDA) in a rabbit myocyte after blocking Na-Ca exchange current by SEA0400 to promote negative Ca-to-APD coupling.¹⁰⁹ C, Theoretical stability boundary of alternans for positive Ca-to-APD coupling.¹²³ Arrows along both axes indicate the directions of increasing instability and dashed lines are the stability boundaries (from stable to unstable) when Ca and APD are not coupled. Solid line is the stability boundary when Ca and APD are coupled, below which the system is stable without alternans and above which alternans occurs in the form of EMCA. D, Same as C but for negative Ca-to-APD coupling. The regions of EMCA, EMDA, and electromechanically quasiperiodic alternans (EMQA) are marked. E, Quasiperiodic APD alternans caused by infusion of erythromycin in guinea pig hearts.¹¹¹

Conversely, APD also directly influences the CaT. By favoring SR Ca reuptake over Ca extrusion via electrogenic Na-Ca exchange current, a longer APD tends to increase SR Ca loading and augment the subsequent CaT. However, at constant heart rate, a longer APD also results in a shorter DI, which can impinge on the recovery of LCCs and RyRs available to release SR Ca on the next beat, thereby reducing the subsequent CaT. Thus, lengthening APD in the present beat can either enhance Ca release or reduce Ca release on the following beat, causing APD-to-Ca coupling to be positive or negative. Generally, during steep APDR slope-induced alternans, APD-to-Ca coupling is positive, but negative coupling can occur at slow heart rates due to SR Ca loading effects.¹⁰⁶ During early repolarization-induced alternans, APD-to-Ca coupling is also positive, since early repolarization truncates Ca entry and release during the very short action potential, after which the long DI allows full recovery of LCC and RyR excitability. For EAD-induced alternans, however, APD-to-Ca coupling tends to be negative, since the prolongation of APD by EADs increases the SR Ca load available for release on the subsequent beat with the shorter APD.

Theoretical analysis^56,110 shows that when Ca-to-APD and APD-to-Ca coupling are both positive, the voltage-driven and Ca-driven instabilities work synergistically to cause the system to become more unstable (ie, more prone to alternans; Figure 5C). Conversely, when the Ca-to-APD coupling and APD-to-Ca coupling have opposite signs, the antagonism makes the system more stable (ie, less prone to alternans; Figure 5D), but eventually exhibits either electromechanically concordant or discordant alternans depending on which instability predominates. If the instabilities of the Ca subsystem and voltage subsystem are evenly balanced, however, a new behavior emerges in which the APD and CaT alternans oscillates quasiperiodically (the EMQA region in Figure 5D). Although direct experimental demonstration of the quasiperiodic behavior is not available, the quasiperiodic APD alternans caused by infusion of erythromycin in guinea pig hearts (Figure 5E)¹¹¹ can potentially be explained by this mechanism.

Tissue-Scale Manifestations of Alternans

So far, we have outlined the features of alternans at the cellular scale. We now turn to the tissue scale in which adjacent myocytes are coupled via gap junctions allowing voltage but not intracellular Ca to diffuse efficiently from myocyte-to-myocyte.

Spatially Discordant Alternans

APD alternans that is uniformly in-phase throughout the whole tissue is called spatially concordant alternans (Figure 6A) (not to be confused with electromechanically concordant alternans). However, APD alternans can also become out-of-phase in different regions of the tissue, called spatially discordant alternans (SDA) (Figure 6B).^112–114 In this case, the out-of-phase regions are separated by a nodal line in which APD is constant from beat to beat. SDA is a highly arrhythmogenic situation because it generates marked dispersion of refractoriness that can result in unidirectional conduction block and initiation of reentry when the tissue APD gradient across the border of the discordant regions reaches a critical steepness, as discussed in detail later.

Figure 6.

Spatially discordant alternans (SDA) and regional spatiotemporal chaos in cardiac tissue. A, At a critical heart rate, action potential duration (APD) begins to alternate concordantly in a long-short pattern over the entire simulated 2-dimensional homogeneous tissue (spatially concordant alternans). During either the long or short APD beat, the APD gradient is minimal, but the QT interval alternates in the simulated electrocardiogram (ECG) tracing below. B, As heart rate increases sufficiently to encounter conduction velocity restitution (CVR), APD alternans becomes spatially discordant, exhibiting a short-long pattern in region a and a long-short pattern in region b, separated by a nodal line without alternans (white line). The result is a marked APD gradient which reverses on alternating beats, causing both T-wave alternans (TWA) and QRS alternans (QRSA) in the simulated ECG tracing below. C, Attenuation of the amplitude of APD alternans (Δa) by spatially dyssynchronous calcium transient (CaT) alternans (Δc_i) in a simulated 1-dimensional cable of rabbit ventricular myocytes. APD alternans amplitude is greater when CaT alternans is synchronized in large clusters of adjacent myocytes (left) compared with a more random spatial distribution (right).¹²³ D, Regional spatiotemporal chaos in a simulated 1-dimensional cable containing a long APD region at the top (red) next to a region with chaotic early afterdepolarizations (EAD) at the bottom (black).⁴⁷ The APD variation in the chaotic region causes the repolarization sequence to vary from beat to beat such that the T-wave amplitude, shape and polarity changes chaotically from beat to beat, causing marked T-wave lability, even though the QT interval (determined primarily by the long APD region) remains unchanged.

Several mechanisms have been shown to cause SDA. Conduction velocity restitution (CVR) is 1 key mechanism.^114,115 Analogous to APDR, the CVR curve quantifies the decrease in conduction velocity as the DI shortens due to incomplete recovery of the Na current. In this experimentally well-supported mechanism,^116–120 the onset of SDA occurs at heart rates which encounter CVR (ie, in which the DI is short enough to slow conduction velocity due to CVR), which typically occurs at fast heart rates. However, other experimental^120–122 as well as simulation and theoretical studies^32,123 have shown that SDA can also be induced by premature ventricular extrasystoles (PVC) and tissue heterogeneities without encountering CVR. Since encountering CVR is not required, these latter mechanisms can produce SDA at normal or even slow heart rates.

Regional alternans can also result in SDA-like patterns, when alternans amplitude varies significantly from 1 region to another, thereby generating dispersion of repolarization. An extreme case is regional 2:1 block due to APD dispersion or weak excitability.

Unlike rapid diffusion of voltage from cell-to-cell through gap junctions, intercellular Ca diffusion is slow and limited¹²⁴ such that the CaT can alternate out-of-phase in neighboring cells. However, if CaT alternans is out-of-phase over small spatial scales, the APD differences will be smoothed by voltage diffusion and diminished (Figure 6C, right). Only if CaT alternans is synchronized over large spatial clusters can a large APD gradient form in response (Figure 6C, left). Recent theoretical analysis¹²³ has shown that maintained synchronization is promoted by large initially synchronized spatial Ca SDA clusters or when the APD-to-Ca coupling is strong. When APD and CaT alternans are dyssynchronous, their nodal lines may or may not co-localize with each other in tissue, as also noted in experiments.^{36,116,121,122}

Regional Chaos Synchronization Causing T-Wave Lability

Unlike SDA generated by either CVR or pre-existing repolarization heterogeneities, when APD alternans transitions into chaotic APD behavior (Figure 2), voltage diffusion will cause tissue APD to spontaneously synchronize regionally, but not globally, such that the regional APD distribution varies chaotically from beat to beat.⁴⁴ This will cause the repolarization sequence and hence the T-wave morphology to vary irregularly from beat to beat, resulting in T-wave lability (Figure 6D).⁴⁷ The underlying marked dispersion of repolarization creates a highly arrhythmogenic tissue substrate susceptible to initiation of reentry by not only exogenous PVCs, but also by self-generated PVCs, as described in the next section.

Alternans and Arrhythmogenesis

We previously classified ventricular arrhythmias into 3 dynamics-based categories³⁰: Ca cycling disorders, prolonged repolarization disorders, and early repolarization disorders. As discussed above, cardiac alternans is a common feature of all 3, and, in the form of SDA or regional chaos synchronization, results in marked dispersion of repolarization, creating a tissue substrate highly vulnerable to reentry. In this setting, a PVC arising from a different location or an exogenous source that propagates into a region with a steep APD gradient can block locally, resulting in 2 wavefronts that propagate around the long APD region to form a figure-of-eight reentry, classically known as the R-on-T phenomenon (Figure 7A and 7B). Moreover, simulations and experiments also indicate the APD gradient can itself spontaneously generate a PVC during its repolarization process that then propagates around the long APD region to initiate reentry (Figure 7C).^125,126 To distinguish between these 2 variations of the R-on-T phenomenon, we refer to the exogenous PVC case as the R-to-T mechanism, and the self-generated PVC case as the R-from-T mechanism. Here we review the conditions under which cardiac alternans leads to initiation of reentrant arrhythmias via these 2 mechanisms. The important conclusion is that cardiac alternans represents more than just an early sign of electrical instability in cardiac tissue, but also serves as a self-contained causal mechanism initiating lethal reentrant ventricular arrhythmias. Although less thoroughly characterized, these findings are also highly relevant to atrial arrhythmias, as discussed in this section later.

Figure 7.

R-to-T and R-from-T mechanisms linking T-wave alternans (TWA) to arrhythmogenesis.

A, Electrocardiogram showing TWA and an R-on-T event (*) initiating Torsade de pointes. B, Voltage snapshot illustrating the “R-to-T” mechanism in which a premature ventricular complex (PVC) (*) emerging from a separate location (lower corner) propagates towards a central prolonged action potential duration (APD) region that is still repolarizing, where it blocks locally and then proceeds around and reenters the blocked region after it has repolarized (dashed arrows), thereby initiating figure-of-eight reentry. C, Voltage snapshot showing the “R-from-T” mechanism in which the PVC (*) is generated directly by electronic current flow from the central region with delayed repolarization into already repolarized tissue. The PVC then conducts around and reenter the heterogeneous region to initiate figure-of-eight reentry (dashed arrows). This requires the central region to have a longer APD (eg, due to an early afterdepolarization [EAD] or action potential [AP] dome) which repolarizes much later than surrounding tissue (eg, without an EAD or AP dome).

Ca Cycling Disorders

The purest example of an arrhythmogenic Ca cycling disorder is CPVT typically caused by ryanodine and calsequestrin mutations.^127,128 However, disordered Ca cycling is also a prominent feature of heart failure, chronic and acute ischemia, digitalis toxicity and other Ca overload conditions, as well as normal hearts subjected to very rapid heart rates. In normal mammalian heart, experimental studies^36–39,129 have shown that alternans occurs well before APDR slope becomes >1, indicating that Ca-driven alternans predominates in setting the heart rate threshold for the onset of TWA in normal human heart. However, since voltage and Ca are coupled, the APDR slope (even if <1) also contributes to either promoting or suppressing the threshold heart rate for the onset of TWA, depending on whether Ca-to-APD and APD-to-Ca coupling have the same or opposite signs (see Figure 5C).

As pacing rate accelerates, spatially concordant APD and CaT alternans develop first, but this does not markedly increase APD dispersion, since on each beat, APD is either uniformly long or uniformly short (Figure 6A). However, reentry can be initiated when loss of 1:1 capture occurs or a properly-timed PVC occurs, assuming some pre-existing APD dispersion is present.

With further increases in heart rate sufficient to encounter CVR, spatially concordant alternans transitions to SDA in which APD dispersion markedly increases (Figure 6B). In this setting, a PVC emerging from the short APD region can block locally as it attempts to propagate across the nodal line where the APD gradient is steepest, proceed laterally to circumvent the line of conduction block and then reenter the long APD region after it recovers excitability to initiate reentry. However, since all real cardiac tissue is inherently heterogeneous, such that APD alternans amplitude is larger in some regions than others, an exogenous PVC may not be required. Instead, a paced beat can develop localized conduction block in the region where the APD gradient is steepest during SDA. In this setting, SDA alone, without a PVC arising from an independent source, is causally sufficient to induce reentry. The same mechanisms apply to transient SDA induced by a PVC or regional tissue heterogeneity at DIs not short enough to encounter CVR.

In normal ventricles, the initiation of lethal arrhythmias by the SDA mechanism requires nonphysiologically rapid heart rates, generally >250–300 bpm in human hearts. However, disease conditions that impair Ca cycling at physiological heart rates, such as CPVT, heart failure, acute and chronic ischemia, digitalis toxicity, can cause the same lethal arrhythmias to occur at much slower heart rates. Moreover, when Ca cycling is impaired, ventricular tissue not only becomes predisposed to Ca-driven alternans and SDA, but also to delayed afterdepolarizations (DAD) and triggered activity caused by spontaneous diastolic SR Ca release. Thus, even if SDA creates a vulnerable substrate that does not generate unidirectional conduction block and initiate reentry on its own, Ca cycling disorders also endogenously generate DAD-mediated PVC triggers as well as dispersion of excitability that can promote reentry initiation.¹³⁰

Prolonged Repolarization Disorders

The purest examples of arrhythmogenic prolonged repolarization disorders are inherited or drug-induced LQTS. Prolonged repolarization is also a feature of heart failure and chronic ischemic heart disease in which electrical modeling reduces repolarization reserve. TWA, in this setting, is mainly voltage-driven by EAD-induced alternans, which occurs when heart rate slows sufficiently to allow EADs to emerge on every other beat, as illustrated in Figure 2B. Within the span of heart rates causing alternans, however, there are typically also windows in which higher order periodicities and frankly chaotic behavior emerge. In these chaotic windows, regional chaos synchronization⁴⁷ causes the heart tissue to form regions with prolonged APD due to EADs next to regions with shorter APD without EADs, resulting in marked regional APD dispersion. Therefore, in this setting, large APD gradients can result from regional alternans,⁴⁷ non-CVR-mediated SDA,³² or regional chaos synchronization.⁴⁴ Not only is the resulting substrate vulnerable to initiation of reentrant arrhythmias by an exogenous PVC, but the substrate can also directly generate a PVC to initiate reentry at a distant site by the R-to-T mechanism (Figure 7B), or locally at the same site by the R-from-T mechanism (Figure 7C).^47,125,126 Electrocardiographically, the chaotic shifting of the EAD regions on each beat when regional chaos synchronization occurs causes the T-wave morphology to vary from beat to beat, resulting in T-wave lability (Figure 6D).

As mentioned earlier, APD alternans can also occur at normal or slow heart rates via the steep APDR slope-driven mechanism when no EADs are present, as long as APD is sufficiently prolonged (eg, left in Figure 2A). In this case, regional alternans or CVR-mediated SDA can generate steep enough APD gradients in LQTS to initiate reentry via either of the 2 mechanisms in Figure 7.

Early Repolarization Disorders

The purest examples of arrhythmogenic early repolarization disorders are Brugada syndrome and short QT syndromes.¹³¹ However, early repolarization is also a feature of acute myocardial ischemia in which activation of ATP-sensitive K channels and other metabolic sequelae enhance repolarization reserve.⁴⁹ In these settings, TWA is mainly voltage-driven by early repolarization-induced alternans, occurring at heart rates in which all-or-none early repolarization occurs on every other beat (Figure 2C). As in LQTS, however, within the span of heart rates causing alternans are windows in which higher order periodicities and frankly chaotic behavior emerge. In these chaotic windows, regional chaos synchronization⁴⁷ can cause the heart tissue to form early repolarizing regions with very short spike-only action potentials next to regions with normal spike-and-dome action potentials, resulting in marked tissue APD dispersion. Thus, 3 separate mechanisms can cause large APD gradients in this setting: regional chaos synchronization,⁴⁴ regional variation in alternans amplitude,⁴⁷ and non-CVR-mediated SDA.³² The resulting marked APD dispersion makes the tissue susceptible to initiation of reentry by an exogenous PVC via the R-to-T mechanism (Figure 7B). However, the steep repolarization gradient can also self-generate a PVC, due to electronic current flow between a still-depolarized spike-and-dome action potential region and an adjacent spike-only action potential region that has repolarized early and can be reexcited. This reexcitation results in so-called phase-2 reentry¹³² via the R-from-T mechanism, as illustrated in Figure 7C. In Brugada syndrome, TWA has been widely observed^133–136 and often occurs simultaneously with PVCs,^133,136 consistent with the R-from-T mechanism. However, the role of early repolarization and phase-2 reentry in Brugada syndrome (the so-called repolarization hypothesis) is currently controversial. Some investigators favor an alternative depolarization hypothesis that structural changes due to fibrosis in regions of the right ventricular epicardium cause slow conduction predisposing these patients to reentry initiation.^137,138 The 2 hypotheses are not necessarily mutually exclusive, however, and could be synergistic or even causally related.¹³⁹ Overall, the role of TWA as a predictor of serious arrhythmia events in Brugada syndrome remains controversial.^134,135

Atrial Arrhythmias

Atrial repolarization alternans, including SDA, has been detected during sinus rhythm or rapid pacing immediately prior to the onset of atrial fibrillation (AF) in both humans^38,140,141 and animal experiments.^142–145 The mechanisms of APD and CaT alternans have been analyzed in experimental and simulation studies of both isolated atrial myocytes and atrial tissue,^{61,83,85,89,145–149} with most of these studies suggesting that the onset of alternans is primarily Ca-driven due to the RyR refractoriness mechanism.^85,89,148 For example, in a clinical study by Narayan et al,³⁸ patients with a history of persistent AF typically developed APD alternans during atrial pacing at near-normal heart rates at which APDR slope was <1, pointing to Ca-driven alternans related to AF-induced Ca cycling remodeling as the predominant initiating mechanism. In contrast, control patients required faster pacing rates to develop APD alternans, but its onset still bore no consistent relationship to APDR slope being >1 or <1. This agrees with the observation by Pearman et al¹⁴⁷ in sheep atria that APD and CaT alternans at slow heart rates were abolished when SR Ca cycling was disabled by thapsigargin treatment, but not at faster heart rates at which ADPR slope was steeper. A recent comprehensive study in perfused canine atria by Liu et al⁸⁹ came to similar conclusions, and also documented with optical mapping that wavebreak causing AF initiation occurred after the transition from spatially concordant alternans to SDA as pacing rate increased, as had previously been observed in humans.¹⁴¹ CaT alternans in atrial tissue is likely to be facilitated by the sparse T-tubular network in atrial myocytes,^150–152 which steepens the CaT restitution curve by facilitating spark-induced spark recruitment.^82,92 To summarize, both voltage-driven and Ca-driven mechanisms of alternans are important in atrial tissue. The coupling between voltage and Ca implies that even when the APDR slope is <1, voltage-driven mechanisms are contributing to either promoting or suppressing the onset of TWA by Ca-driven mechanisms, depending on whether Ca-to-APD and APD-to-Ca coupling have the same or opposite signs (see Figure 5C).

Other voltage-driven alternans mechanisms may also play a role in atrial arrhythmogenesis, including AF. Since EADs have been shown to occur in atrial myocytes,^153,154 in principle, EAD-induced alternans and regional chaos synchronization initiating reentry by mechanisms equivalent to R-from-T or R-to-T could also occur in atria. In aged rats, for example, mild glycolytic inhibition induced EADs alternans followed by rapid EAD-mediated triggered activity in the pulmonary veins that initiated AF (see Figure 5 in Ono et al¹⁵⁵). In this case, rapid triggered activity from the pulmonary veins was most likely fast enough to induce SDA in the body of the atria which resulted in wavebreak initiating AF. Whether regional chaos synchronization played any role is unclear.

Whether the early repolarization alternans mechanism can cause atrial alternans and regional chaos synchronization to initiate AF is currently unclear. Due to its large I_to conductance, the atrial action potential is typically spike-like without a prominent action potential dome. Rapid atrial pacing, however, is well known to induce electrical remodeling that downregulates I_to. It is intriguing to speculate if I_to downregulation was sufficient to allow an action potential dome to reemerge intermittently in atrial myocytes, then the early repolarization alternans mechanism, as well as regional chaos synchronization and phase-2 reentry, could come into play. As yet, however, there is no experimental evidence directly supporting this conjecture.

Focal Arrhythmias and Anatomic Reentry

Cardiac alternans plays less of a direct mechanistic role in focal arrhythmias, with the possible exception of bidirectional VT seen in Ca cycling disorders like CPVT and digitalis toxicity. A mechanism for bidirectional VT proposed from simulations is DAD alternans, in which alternating sites of DAD-mediated triggered activity in the different branches of the His-Purkinje system or regions of ventricular myocardium trigger each other reciprocally in a ping-pong fashion such that the QRS morphology alternates, with secondary TWA changes due to the altered repolarization sequence.¹⁵⁶ If the rate of bidirectional VT is fast enough to induce SDA, however, the resulting wavebreaks can cause degeneration into ventricular fibrillation, just like rapid-pacing induced VT or ventricular fibrillation.

For anatomic VT due to reentry around an anatomic obstacle, as occurs in scar-related VT and bundle branch macro-reentry, TWA and QRSA are also less important. However, transient cycle length alternans, related to CVR in slowly-conducting regions of the reentry circuit, is commonly observed, especially during initiation and termination of reentry. Transient cycle length alternans is also commonly observed in supraventricular arrhythmias involving anatomic reentry, including atrio-ventricular nodal reentry, and atrio-ventricular reentry using an accessory pathway(s).

Back to the Bedside

Electrocardiogram Signatures of Alternans

The electrocardiogram represents a weighted average of the voltage differences within the heart during the cardiac cycle, with the amplitude, shape, polarity and duration of various electrocardiogram components reflecting the underlying timing, direction and synchronicity of propagating depolarization/repolarization waves. Here we discuss the links between the cellular/tissue-scale mechanisms of alternans and their corresponding electrocardiogram patterns. For this purpose, we equate the QT and TQ (time interval between the T-wave and the next Q-wave) intervals (see Figure 8A) with average estimates of tissue APD and DI, respectively. Thus, during alternans at a constant heart rate, just as APD and DI alternate oppositely (i.e. long-short and short-long, respectively) such that pacing cycle length(=APD+DI) remains constant for each beat, the QT and TQ intervals also alternate oppositely such that RR = QT+TQ for each beat.

Figure 8.

Clinical examples of alternans on recorded electrocardiograms. A, T-wave alternans (TWA) with long TQ (time interval between the T-wave and the next Q-wave) intervals in a 52-year-old man with amiodarone-induced long QT syndrome (LQTS).²⁴ RR=QT+TQ. B, TWA with short TQ intervals from a 13-month-old girl with congenital LQTS.¹⁵⁷ C, TWA with T-wave polarity alternans (TWPA) and QRS complex (QRSA).¹⁵⁸ D, TWA with TWPA without QRSA in a 59-year old man with LQTS.¹⁶³ E, QRSA without TWA. F, Pulsus alternans without TWA.¹⁶⁸ G, Microvolt TWA detected using frequency-domain analysis (right).¹⁷⁰ H, A 1 year-old girl with Jervell-Lange-Nielson syndrome with multiple episodes of T-wave lability and Torsade de Pointes, who later died suddenly at 2 years of age.¹⁷¹ FFT indicates fast fourier transform.

In simulated homogeneous tissue in which spatially concordant APD alternans is present with constant conduction velocity, the spatial repolarization sequence will match the depolarization sequence such that the amplitude, shape, and polarity of both the QRS and T-wave will be nearly identical for both long and short QT beats (Figure 6A). However, when the TQ interval (surrogate for DI) becomes short enough to encounter CVR and cause SDA, the changing conduction velocity of the depolarization wave as it propagates through the tissue will reverse on alternate beats, causing QRSA (Figure 6B).^113,114 Likewise, the out-of-phase alternating APD regions will increase the dispersion of repolarization and cause its spatial sequence to reverse on alternate beats, such that the T-wave widens and reverses polarity on alternate beats even in initially homogeneous tissue.

In real heart tissue, however, regional APD heterogeneity exists even in normal ventricles due to transmural and base-to-apex APD gradients, and is often exacerbated by heart diseases. As a result, the spatial repolarization sequence and hence the T-wave amplitude, shape, and polarity often change on alternate beats, even during spatially concordant alternans. During SDA, however, the alternans in T-wave amplitude, shape, and especially its polarity becomes even more prominent. If SDA results from encountering CVR at short TQ intervals, then the QRS also alternates. If the TQ intervals are too long to encounter CVR and QRSA is therefore absent, then some other mechanism must be operative for SDA to occur. In principle, CaT alternans due to alternating SR Ca load can produce TWA at long TQ intervals. However, when alternans is Ca-driven, the TWA amplitude may not be as prominent if Ca-to-APD coupling is weak or if CaT alternans is spatially dyssynchronous.

Illustrative electrocardiogram traces from real patients are shown in Figure 8.

TWA with and without QRSA

In Figure 8A, the electrocardiogram trace from a patient with amiodarone-induced LQTS illustrates TWA with 2 relatively long alternating TQ intervals.²⁴ These long TQ intervals could plausibly correspond to DIs positioned on either side of the discontinuous steep region at which EADs suddenly prolong the APD in the APDR curve, as in Figure 2B, causing the QT intervals to alternate by the EAD-induced alternans mechanism. Note that in this tracing, both TQ intervals are too long to encounter CVR and cause SDA, as evidenced by the absence of QRSA. The alternans in T-wave amplitude, shape, and polarity are nevertheless prominent, however, and could potentially indicate underlying SDA due to a mechanism other than encountering CVR.

Figure 8B shows another electrocardiogram tracing from a congenital LQTS patient,¹⁵⁷ in which one of the alternating TQ intervals is potentially short enough to encounter CVR. TQ alternans is also accompanied by prominent alternans in the T-wave amplitude and shape. However, QRSA is not present, indicating that if SDA is present, it is not related to encountering CVR.

Figure 8C, also obtained from an LQTS patient,¹⁵⁸ illustrates TWA with short TQ intervals and prominent T-wave polarity alternans, which is widely observed in LQTS.^{19,22,159–162} Moreover, QRSA is also present, indicating a high probability of SDA due to encountering CVR at the short TQ intervals. The mechanism of alternans could be EAD-induced alternans (Figure 2B), but steep APDR slope-induced alternans (Figure 2A) or Ca-driven alternans (Figure 5) are also plausible in view of the short preceding TQ intervals.

T-wave polarity alternans is also observed in LQTS patients with long TQ intervals and no QRSA (Figure 8D),^159,163 indicating that a non-CVR mechanism of SDA or regional alternans may be responsible. In Holter recordings, macroscopic T-wave polarity alternans episodes have been observed in up to 45% of congenital LQTS,¹⁶⁴ and often precede, sometimes immediately, episodes of Torsade de pointes, as illustrated in Figure 1B.

Early repolarization-mediated alternans in patients with early repolarization syndromes can also exhibit TWA with normal-to-long TQ intervals corresponding to the DIs at which all-or-none early repolarization suddenly occurs, as illustrated in the APDR curve in Figure 2C.

In CPVT and failing human hearts with disordered Ca cycling, Ca-driven alternans, rather than steep APDR slope voltage-driven alternans, is likely to be the mechanism initiating TWA, manifested initially as spatially concordant alternans with pulsus paradoxus. As heart rate increases and TQ intervals become short enough to encounter CVR, alternans transitions to SDA with QRSA. This is also true for normal human heart when paced nonphysiologically at very rapid rates (>250–300 bpm). Animal studies have confirmed that SDA in this setting is one of the key mechanisms by which very rapid pacing initiates reentrant polymorphic VT and ventricular fibrillation in normal hearts.^112,113

Other causes of QRSA, with and without secondary TWA

QRSA, usually with secondary T-wave changes, can also occur in other conditions, such as bidirectional VT seen in Ca cycling disorders like CPVT and digitalis toxicity, and has been hypothesized to arise from DAD alternans.¹⁵⁶ Bundle branch block alternans is another form of QRSA sometimes observed in patients with conduction system disease, attributed to complex patterns of conduction block.¹⁶⁵ QRSA can also be observed with large pericardial effusions, thought to be related to pendulous motion of the heart swinging in the fluid-filled pericardial sack.¹⁶⁶ Finally, QRSA can occur without visible TWA (Figure 8E), although whether subtle microvolt TWA is present has not been excluded (see below).

Pulsus alternans without TWA

Pulsus alternans usually occurs with TWA or QRSA.^157,167 If alternans is Ca-driven and Ca-to-APD coupling is very weak, pulsus alternans can occur without discernible visible TWA^168,169 (Figure 8F).

Microvolt TWA

Microvolt TWA is a subtle form of repolarization alternans characterized by microvolt changes in T-wave amplitude that are not visible to the naked eye, but can be revealed by signal processing using either frequency-domain or time-domain techniques when heart rate is mildly elevated to 90–110 bpm (Figure 8G).^29,170 The exact mechanism underlying microvolt TWA is unknown, but 1 possibility is subtle Ca-driven alternans with weak Ca-to-APD coupling. In this case, mechanical pulsus alternans might be discernible without visibly appreciable TWA on electrocardiogram. Alternatively, even if Ca-to-APD coupling is strong, when CaT alternans is spatially dyssynchronous within or between adjacent myocytes, the alternans in voltage may be attenuated (see Figure 6C), and pulsus alternans also weak or absent. Another possibility is APD alternans or 2:1 conduction block in a small region due to fibrosis or ischemia, with an electrical mass too small to be visibly detected in the global electrocardiogram signal.

Microvolt TWA is common in patients with cardiac diseases and its efficacy for risk stratification of arrhythmias and sudden death has been widely investigated.^13,29 Clinically, microvolt TWA testing has a higher negative predictive value than positive predictive value; that is, a negative test indicates a low probability of an event, whereas a positive test is less informative.²⁹

T-wave lability

Although not as common as TWA, patients with heart disease can also exhibit T-wave lability, in which T-wave features vary irregularly from beat to beat (Figure 8H).¹⁷¹ This appearance is consistent with the regional chaos synchronization mechanisms that come into play when dynamical chaos is present (Figure 6D).^44,47 For prolonged repolarization-induced arrhythmias, this occurs when the heart rate encounters the steep discontinuous region in the APDR curve at which EADs behave chaotically as illustrated in Figure 2B, causing regions with EADs to appear whose location in the tissue shifts irregularly from beat to beat (Figure 6D) such that the spatial repolarization sequence defining the features of each T-wave are different. The resulting dispersion of repolarization makes the tissue highly susceptible to initiation of reentry by both R-from-T and R-to-T mechanisms,^44,47 supporting the clinical observations of a high association between T-wave lability and Torsade de pointes episodes.^171–173

A similar situation may occur with early repolarization-mediated arrhythmias, due to regional chaos synchronization when the heart rate transitions from alternans into one of the nearby chaotic windows (Figure 2C). On the electrocardiogram, this would also manifest itself as T-wave lability reflecting marked underlying dispersion of repolarization from beat to beat, creating a vulnerable substrate with increased susceptibility to phase-2 reentry initiated by the R-from-T mechanism.

In CPVT, as well as normal and failing hearts, Ca-driven alternans or steep APDR slope-induced alternans can induce chaotic APD behavior when loss of 1:1 conduction occurs during sinus rhythm or non-reentrant VT, resulting in complex repolarization patterns and T-wave dynamics that facilitate initiation of reentry.

In summary, although T-wave lability is not as prevalent as TWA, when it occurs, it is even more arrhythmogenic due to the chaotic repolarization dynamics.

P-wave alternans

P-wave alternans, possibly the atrial equivalent of QRSA, has been observed only rarely in humans, and not clearly associated with atrial flutter or AF.¹⁷⁴ Atrial repolarization is not visible on a standard electrocardiogram, but atrial repolarization alternans has been detected during sinus rhythm or rapid pacing prior to AF initiation using intracardiac catheters in both humans^38,140,141 and animal experiments.^142–145

Therapeutic Implications

With cardiac alternans serving both as an indicator of cardiac dynamical instability underlying high arrhythmia risk, as well as a direct causal mechanism that can initiate potentially lethal ventricular arrhythmias, targeting suppression of cardiac alternans pharmacologically or by other means is a promising antiarrhythmic strategy. However, with 3 categories of life-threatening arrhythmias to contend with, 5 underlying voltage- and Ca-driven mechanisms of alternans, and disease conditions that often involve multiple dynamic arrhythmia categories as well as multiple alternans mechanisms, the task of developing an effective strategy is not straightforward. In addition, lethal cardiac arrhythmias are also precipitated by other factors besides alternans, and interventions designed to suppress cardiac alternans must not exacerbate those other mechanisms of arrhythmogenesis. That said, the effects of some common antiarrhythmic drugs on cardiac alternans are summarized below.

Beta adrenergic blockers

For arrhythmias potentiated by adrenergic stimulation, such as CPVT, acute and chronic ischemia, some LQTS subtypes, beta blockers have multiple antiarrhythmic effects. These include: (1) slowing heart rate to avoid very short DIs that promote steep APDR slope-induced alternans, Ca-induced alternans and CVR-mediated SDA; (2) blunting adrenergic stimulation of LCCs that promotes EADs, EAD-mediated alternans and chaotic APD dispersion promoting reentry initiation; (3) blunting adrenergic stimulation of SERCA pumps that increases SR Ca loading into the steep region of the fractional SR Ca release curve that promotes CaT alternans; (4) blunting adrenergic enhancement of RyR excitability and leakiness that promote Ca waves and DAD-mediated triggered activity; (5) reducing cardiac workload to attenuate the severity of acute myocardial ischemic episodes that can lead to early repolarization-induced alternans and phase-2 reentry. Generally, these effects make beta blockers safe as first-line antiarrhythmic therapy.

Na channel blockers

Na channel blockers decrease the amplitude and slow the recovery of the Na current,¹⁷⁵ altering both APDR and CVR such that alternans and SDA occur at slower heart rates,¹⁷⁶ particularly in the setting of acute and chronic ischemic heart disease in which the Na channel properties are already similarly altered by electrical remodeling.¹⁷⁷ This may have been an important factor contributing to the increased mortality in patients with ischemic heart disease treated with Na channel blockers in the Cardiac Arrhythmia Suppression Trial (CAST).¹⁷⁸ Blocking Na channels can promote early repolarization-induced alternans, which may contribute to inducing TWA in Brugada syndrome.^{135,179–181} A more promising class of Na channel blockers, however, may be selective late Na channel blockers that shorten the QT interval, abolish TWA, and suppress arrhythmias in patients with LQTS.^21,182

Ca channel blockers

High doses of Ca channel blockers shorten APD and flatten APDR slope, and thus can suppress steep APDR slope-induced alternans. Ca channel blockers also suppress EADs and EAD-induced alternans. However, the high dosages required often suppress contractility too severely to be useful clinically. Similar to Na channel blockers, Ca channel blockers can promote early repolarization-induced alternans and phase-2 reentry in Brugada syndrome. Blocking Ca channels can either promote or suppress Ca-driven alternans depending on the setting,^63,80,92 as has been observed with verapamil.^{141,183–185} Drugs that block the Ca window current selectively without reducing its peak amplitude, although not yet clinically available, have been shown to suppress EADs and prolonged repolarization-induced alternans without unduly suppressing the CaT and contractility.^{43,48,125,186} They may also be effective at suppressing early repolarization-induced alternans and the early repolarization-induced instability.¹⁸⁷

K channel blockers and openers

K channel blockers lengthen APD and steepen APDR slope, promoting steep APDR slope-induced alternans. K channel blockers also promote EADs and EAD-induced alternans, which may have contributed to the increased mortality in patients with ischemic heart disease treated with d-sotalol in the Survival With ORal D-sotalol (SWORD) trial.¹⁸⁸ For early repolarization-induced alternans, blocking K currents, particularly I_to (eg, with low dose quinidine), suppresses alternans. The effects of K channel blockers on Ca-driven alternans are still unclear. K channel openers have been shown to suppress voltage-driven alternans by flattening APDR slope¹⁸⁹ and suppressing EADs in the setting of LQTS.^190,191

SERCA overexpression

It has been shown that overexpressing SERCA suppresses both mechanisms of Ca-driven alternans.^81,96 However, increasing SERCA activity can overload the SR, which increases the susceptibility to Ca waves and thus DAD-induced triggered activity.

Summary

We have reviewed how an astute clinical observation was the starting point for a 150 year bedside-to-bench-and-back-again journey involving exemplary collaborative science by multiple generations clinicians, experimental and theoretical biologists. Unraveling cardiac alternans has been a vital step leading to deeper understanding of the dynamics of cardiac arrhythmias that can end the lives of patients across a wide spectrum inherited and acquired human cardiovascular diseases. It is complicated, with 3 voltage-driven and 2 Ca-driven mechanisms that serve as warning signs that the heart is nearing the threshold for more complex instabilities that can degenerate into serious ventricular and atrial arrhythmias. As such, alternans is a common precursor to all 3 of the major categories of cardiac arrhythmias, including disorders of Ca cycling, prolonged repolarization, and early repolarization, spanning inherited diseases such as CPVT, LQTS, short QT syndromes, and Brugada syndrome to common diseases such as acute and chronic myocardial ischemia, heart failure, and drug toxicity. Not only is cardiac alternans an early warning sign of more life-threatening cardiac instabilities to come, but also plays a direct causal role in initiating reentry, by creating a vulnerable substrate for reentry initiation by exogenous PVCs arising via the R-to-T mechanism or by PVCs self-generated by the repolarization gradients via the R-from-T mechanism. The multifactorial nature of cardiac alternans and its involvement in multiple arrhythmia mechanisms make it a daunting antiarrhythmic target, but a highly worthwhile goal for the upcoming generation of scientists and clinicians to pursue in the same collaborative spirit.

Article Information

Acknowledgments

We dedicate this review article to the late David S. Rosenbaum, MD, a consummate physician-scientist who made seminal research contributions to both the bench and bedside aspects of cardiac alternans. This study was supported by National Institutes of Health grants R01 HL134709, R01 HL139829, R01 HL134346, and P01 HL078931.

Disclosures

None.

Sources of Funding

None.